Neoepitopes arise from somatic mutations in the tumor

Data generation from mouse tumor models

  • Two cell lines (CT26 colon carcinoma and 4T1 breast cancer)
  • Two organs (spleen and tumors)
  • With and without checkpoint inhibitor treatment

Method

Results

Response filtering

  • Response selection criteria: logfold change > 2 and p-value < 0.01

Expression level and rank score

  • Now selection criteria is eluted ligand rank < 2% and expression > 0.1 TPM - top 500 neoepitopes
  • Can these selection critereia be optimized

Elution, binding affinity rank scores and self-similarity

Improved binding affinity

Mutation position

Modelling

  • Sampling 40 negative and the 30 positive.

R package and shiny app

We have developed barcc package

And a shiny app for rapid visualization of the data

Discussion

  • Data set is to small to see a clear pattern in immnunugenic and non-immunugenic neoepitopes
  • Also shown by a poor performance of the model
  • But! With a trend that if we take into account multiple neoepitope characteristics, we can improve immunogenic selection

Further work:

  • Try out with more data, when more peptides are screened
  • Is the results the same for human data
  • Make packages for human data

Thank you!